art B2. Istanbul 2004 
at 20 m/s (airspeed) is 
eds. In extreme cases, 
height to avoid being 
ude; above that height, 
; air traffic control, and 
vs for efficient mission 
68 W/m? of energy is 
|). Due to atmospheric 
ne of year and the 
:d at lower altitudes. 
ce (in Wh/m°/day) as a 
orizontal surface at 55° 
1e UAV needs at least 
yr 7 months (March — 
e 
QV 
N 
e 
E E 
Dec | 
E 
Oct 
  
? ~ . ~ 
17/day) as a function of 
cation at 55° N latitude 
fuel cells are charged 
ent energy during night 
ds, using some of its 
ns during winter time, 
erative fuel cells and a 
short (several days) 
m is always available 
ig large areas (100 000 
taking cloud cover into 
take advantage of even 
r. Indeed, at a modest 
transferred over large 
, even if there are no 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
3. INSTRUMENTS 
3.1 Design philosophy 
The constraints that are imposed onto the design of the remote 
sensing instruments and the auxiliary systems are weight, power 
consumption and volume. At this stage of the project, the UAV 
cannot carry payloads heavier than 2 kg, unless it is scaled up. 
The power available for the payload is of the order of 1 kW, the 
remaining electrical power generated by the solar cells being 
used for the flight systems itself. The volume constraint is 
probably the most difficult one to deal with : the instruments 
should be designed to fit within the limited volume and 
irregular shape of the aircraft’s fuselage. 
Using proven technology and designing the instruments so that 
they conform to but not exceed the applications’ requirements 
allows the development to be time and cost effective. Also, this 
implies that the instruments will deliver data that are “just good 
enough”, reducing significantly the cost of data processing. 
The instruments will be defined and developed in sequence, 
allowing a possible response to changes in the market. In a few 
years from now, the UAV platform will evolve, being able to 
carry heavier payloads and provide more power. 
3.2 Auxiliary payload 
The auxiliary payload is shared between all instruments: a 
GPS/INS system for position and attitude determination and a 
data transmission system (S- or X-band, 75 Mbps). 
For navigation, C/A based GPS positioning is sufficient. 
Attitude determination in real time is only required to support 
the image acquisition : it is used to control the line acquisition 
rate (so that sufficient forward overlap is guaranteed). 
3.3 Implementation time line 
As previously mentioned, the instruments to be carried by the 
UAV will be developed sequentially. Table 3 shows the 
projected time schedule. The first two instruments are currently 
under development, the latter two will be developed from mid- 
2005 onwards. Up to that time, changes in the specifications are 
possible, according to the market requirements. 
2004 2005 2006 2007 
Multispectral digital camera 
LIDAR 
Thermal digital camera 
SAR 
  
Table 3. Sequential development of the instruments. 
3.4 Multispectral Digital Camera 
The Multispectral Digital Camera is the first instrument to be 
implemented. It will provide images in up to 10 narrow spectral 
bands in the visual and near-infrared spectrum (400 — 1000 nm, 
10 nm individual band width), at 15 to 20 cm ground pixel size. 
Due to the multispectral character of the instrument, it is 
implemented as a push broom system, using 12000 pixel wide 
line CCD arrays (see Reulke 2003 for an overview of available 
sensor technology). This results in a swath width of 1800 to 
UJ 
2400 m. Because of the small field of view of the system (6°), 
the images are not suited for stereoplotting, but they are much 
less affected by atmospheric refraction than commercial aerial 
survey systems; furthermore, the effects of the central 
perspective are very limited, making the images well suited for 
orthophoto production. 
Using as high-grade position and orientation system, forward 
oversampling and the use of ground control, a position accuracy 
of 15 cm can be guaranteed. Oversampling is possible because 
of the low air speed of the system. 
The design is based on the worst case situation : 8 hours usable 
for acquisition at equinox at 55? latitude, so that more than 8 
hours can be used for data acquisition during the summer 
months. It is expected to obtain a system signal-to-noise ratio of 
200 (worst case), which is comparable to scanned aerial film (in 
optimum circumstances). The signal will be digitized at 10 bits. 
The system MTF shall be better than 15%, where 10 % is 
deemed to be acceptable. 
Using 7.5m square pixels in the sensor line array (pitch), these 
requirements translate into a focal length of 0.75 m and a lens 
aperture of 0.13 m. This can be realized by a refractive system. 
The expected data-volume produced by the camera is : 12 000 
pixels @ 10 bits per pixel (à) 200 Hz — 22.9 Mbit/s per line 
sensor. When 10 spectral bands are recorded, this results in 229 
Mbit/s. It is clear, however, that these spectral bands are 
correlated, allowing significant data compression prior to 
transmission. 
An 8 hour survey day will yield a total of 0.8 Tbyte of raw data. 
3.5 LIDAR 
The LIDAR instrument will provide elevation information, that 
can be used for orthophoto production of the multispectral 
digital camera and also as information in its own right. 
Covering the same swath as the digicam, it will produce a point 
density between 1 point per 2-4 m^. Even higher point densities 
can be obtained by multiplying overpasses over the same area. 
This could be useful for detailed city mapping (c.g. Noble et al., 
2003). Another application of the high point density is the 
statistical improvement of a DSM/DTM, e.g. for coastal zone or 
flood plane mapping. 
The main challenges in the design of the LIDAR instrument are 
the power that is required for an active instrument and the 
limited mass (5 kg maximum) in which this power has to be 
dissipated. Recently, it has been shown that a LIDAR system 
designed for slant ranges up to 6 km was capable of 
successfully measuring ranges up to 15 km (Haarbrink, 2003). 
To limit the power used by the scan mechanism, a nutating 
mirror setup will be used, so the scan angle (or swath width as a 
function of flying altitude) is fixed. The mirror rotation 
frequency will be constant, too. Together, this will generate a 
quasi-random point distribution. 
The pulse repetition frequency is set to 15 kHz, which will 
produce a point density of 1 point per 2.5 m” "in the best case. 
The instrument will record the first and last reflected pulse, and 
the intensity of the reflected pulses.